Expression of Water-soluble, Ligand-binding Concatameric Extracellular Domains of the Human Neuronal Nicotinic Receptor α4 and β2 Subunits in the Yeast Pichia pastoris

Nicotinic acetylcholine receptors (nAChRs) are ligand-gated cation channels that are responsible for cell communication via the neurotransmitter acetylcholine. The predominant nAChR subtype in the mammalian brain with a high affinity for nicotine is composed of α4 and β2 subunits. This nAChR subtype is responsible for addiction to nicotine and is thought to be implicated in Alzheimer and Parkinson diseases and therefore presents an important target for drug design. In an effort to obtain water-soluble, ligand-binding domains of the human α4β2 nAChR for structural studies, we expressed the extracellular domains (ECDs) of these subunits in the eukaryotic expression system Pichia pastoris. The wild-type ECDs and their mutants containing the more hydrophilic Cys-loop from the snail acetylcholine-binding protein (individually expressed or coexpressed) did not demonstrate any specific interaction with ligands. We then linked the mutated ECDs with the 24-amino acid peptide (AGS)8 and observed that the β2-24-α4 ECD concatamer, but not the α4-24-β2 one, exhibited very satisfactory water solubility and ligand binding properties. The 125I-epibatidine and [3H]nicotine bound to β2-24-α4 with dissociation constants (Kd) of 0.38 and 19 nm, respectively, close to the published values for the intact α4β2 AChR. In addition, 125I-epibatidine binding was blocked by nicotine, cytisine, acetylcholine, and carbamylcholine with inhibition constants (Ki) of 20.64, 3.24, 242, and 2,254 nm, respectively. Interestingly, deglycosylation of the concatamer did not affect its ligand binding properties. Furthermore, the deglycosylated β2-24-α4 ECD existed mainly in monomeric form, thus forming an appropriate material for structural studies and possibly for pharmacological evaluation of novel α4β2 nAChR-specific agonists.

Both the ␣ and ␤ subunits contribute to the pharmacological properties of the binding sites, which have principal and complementary components and lie at the interface between an ␣ (␣2, ␣3, ␣4, or ␣6) and a ␤ (␤2 or ␤4) subunit in the case of heteromeric receptors or between two identical subunits in the case of homomeric receptors. It is thought that homomeric receptors made up of ␣7, ␣8, or ␣9 subunits have five identical acetylcholine-binding sites per receptor molecule (one on each subunit interface), whereas heteromeric receptors including the muscle type receptor have two binding sites per molecule (reviewed in Ref. 7).
The ␣4␤2 nAChR displays the highest affinity for nicotine and constitutes the most abundant type of neuronal nAChR in brain tissues. A stoichiometry of (␣4) 2 (␤2) 3 has been proposed, generating two agonist-binding sites, whereas manipulation of the stoichiometry of ␣4␤2 nAChRs expressed in Xenopus oocytes indicates that (␣4) 3 (␤2) 2 nAChRs can be also viable, displaying higher Ca 2ϩ permeability and a lower affinity for nicotinic agonists (8).
So far, based on the high resolution crystal structures of the acetylcholine-binding protein, a homolog of the extracellular domain of the nAChR found initially in the snail Lymnaea stagnalis (LsAChBP) (15), the Torpedo muscle type receptor (16), the mouse ␣1 ECD (17), and the prokaryotic homologs Gloeobacter violaceus and Erwinia crysanthemi ligand ion channels (GLIC and ELIC) (18,19), homology models for several nAChR subtypes have been generated and present the best available material for drug design (20 -22).
In previous studies, to investigate the pharmacological properties of the human ␣4␤2-nAChR, intact or truncated subunits have been mainly expressed in Xenopus oocytes. Person et al. (23) used this system to express the extracellular domains of the ␣4 and ␤2 subunits with or without the M1 transmembrane domain. They observed that individually expressed or coexpressed water-soluble ␣4 and ␤2 ECDs did not bind ligands and that epibatidine binding was only observed when transmembrane domain M1 was present on at least one of the two coexpressed ECDs. However, such hydrophobic molecules may present the usual difficulties that are encountered in structural studies of membrane proteins. Ideally, large quantities of watersoluble ligand-binding domains of ␣4␤2 AChRs (as well as of other neuronal AChR subtypes), suitable for crystallization, are needed for high resolution structural studies and subsequent rational drug design.
Covalent linkage of specific AChR subunits by flexible linkers has resulted in functional receptors, as described in several studies on ␣4␤2 (24,25) and other AChR subtypes, such as ␣3␤4 (26) and ␣6␤2␤3 (27). More specifically, in the case of ␣4␤2, Zhou et al. (25) expressed intact ␣4 and ␤2 subunits as concatamers, in which the two subunits were linked by an 18 -36-amino acid linker peptide either as ␣4-linker-␤2 or ␤2-linker-␣4. The recombinant receptors had a high ligand binding affinity, and the authors inferred that the (␣4) 2 (␤2) 3 AChR had a higher affinity for epibatidine and a lower Ca 2ϩ influx than the (␣4) 3 (␤2) 2 AChR. Based on this observation, Carbone et al. (24) expressed ␣4␤2 AChR receptors in both stoichiometries as pentameric constructs connected by a flexible linker and proved the hypothesis of Zhou et al. However, the precise native stoichiometry of the molecule is still unclear.
The aim of all of these studies was to elucidate the functional and pharmacological properties of the ␣4␤2 receptor, but, although invaluable for the determination of ␣4␤2 AChR-specific characteristics, they did not lead to the production of large quantities of water-soluble ligand-binding domains of ␣4␤2 AChRs suitable for crystallization and subsequent high resolution structural studies. In fact, to our knowledge, no other ECD of any AChR with a high affinity for small ligands has been constructed, with the exception of the protein ligand ␣-bungarotoxin, the binding of which does not involve a subunit interface and thus binds with high affinity to ␣1 and ␣7 ECDs (28,29).
In this study, we present our efforts aimed at producing soluble ␣4␤2 nAChR domains with a high ligand binding affinity, appropriate for high resolution structural analysis and possibly pharmacological studies. The ECDs of the two subunits were heterologously expressed in the yeast Pichia pastoris in a soluble glycosylated form. Although individual or coexpressed ␣4-and ␤2-ECDs did not bind ligands, we managed to produce a recombinant protein consisting of the linked ␣4and ␤2-ECDs with near native pharmacological properties and excellent water solubility. Because deglycosylation did not alter either the pharmacological properties or the solubility of the protein and because nonglycosylated proteins are more amenable to crystallization than glycosylated ones, we suggest that the deglycosylated form of this protein will be suitable material for crystallization trials.

EXPERIMENTAL PROCEDURES
Cloning of the ␣4and ␤2-ECDs in Expression Vectors-All of the amino acids are numbered according to their position in the mature protein sequence. cDNAs coding for the N-terminal ECDs of the human nicotinic ␣4 (ACHA4_HUMAN, P43681) and ␤2 (ACHB2_HUMAN, P17787) AChR subunits (amino acid residues 1-210 and 1-211, respectively) were enzymatically amplified by PCR from full-length ␣4 and ␤2 cDNAs, respectively, kindly provided by Dr. J. Lindstrom (Department of Neuroscience, University of Pennsylvania Medical School, Philadelphia, PA), using the primer combination ␣4-EcoRI-for/ ␣4-XbaI-rev and ␤2-EcoRI-for/␤2-XbaI-rev. The primers were designed to contain EcoRI (forward) and XbaI (reverse) restriction sites so that the purified PCR products after digestion were subcloned into the EcoRI/XbaI sites of the expression vector pPICZaA (Invitrogen). The sequences of all the oligonucleotides used in this paper are given in supplemental Table S1.
Cloning of the ␣4 and ␤2 Cys-loop Mutant ECDs (mECDs)-The Cys-loop mutant ECDs, in which the base sequence of the native Cys-loop was segmentally replaced by the corresponding sequence of the LsAChBP (ACHP_LYMST, P58154), were generated by two-step PCR using the previously constructed cDNAs as templates. In the case of the ␣4-mECD, two different fragments were generated: (a) first using the primers ␣4-EcoRIfor and ␣4-rev1 and pPICZaA-␣4-ECD cDNA as template, then the primers ␣4-EcoRI-for and rev2-HincII and the product of the first PCR as template, yielding cDNA1 and (b) first using the primers ␣4-for1 and ␣4-XbaI-rev and pPICZaA-␣4-ECD cDNA as template, then the primers for2-HincII and ␣4-XbaI-rev and the product of the first PCR as template, yielding cDNA2. A new HincII restriction site was thus generated in each of the final cDNAs. cDNAs 1 and 2 were then digested, respectively, with EcoRI/HincII or XbaI/HincII, and the product was ligated into EcoRI/XbaI sites in the expression vector pPICZaA (Invitrogen) or a modified pPICZaA with a base sequence encoding a FLAG tag instead of a His 6 tag. ␤2-mECD containing the same restriction sites was constructed in the same way using the primers ␤2-EcoRI-for, ␤2-rev1, rev2-HincII, ␤2-for1, ␤2-XbaI-rev, and for2-HincII.
Yeast Transformation and Screening for Positive Clones-Linearized plasmids were transformed by electroporation into the X33 strain of P. pastoris, and positive clones were selected for small scale cultivation according to the manufacturer's instructions (Invitrogen) and in detail described elsewhere (28,30). The culture supernatants were tested for expression of the recombinant ECDs by dot-blot analysis using anti-Myc 9E.10 mAb (ATCC) or anti-His 6 C terminus antibodies (Invitrogen) and horseradish peroxidase-conjugated polyclonal goat antimouse IgG antibodies (Dako) or using horseradish peroxidaseconjugated anti-FLAG M2 mAb (Sigma), depending on the tag carried by the recombinant protein. The clones with the highest protein yield were used for large scale protein expression.
Protein Purification-The supernatants of cultures containing the secreted ECDs (wild-type or mutant ECDs) were concentrated using a Minitan Ultrafiltration System (Millipore) with a 10-kDa cut-off filter and dialyzed against 20 mM HEPES, 300 mM NaCl, 5% glycerol, pH 8.0, and then the proteins were purified by affinity chromatography on nickel-nitrilotriacetic acid-agarose (Qiagen) or on an anti-FLAG M2 gel (Sigma-Aldrich) depending on the tag used.
For size exclusion chromatography, the purified proteins were concentrated to 500 l by centrifugation (3,000 ϫ g, 30 min, 4°C) in Amicon tubes (Millipore), and applied to an AKTA 90 FPLC purifier system (Amersham Biosciences), using a 24-ml Superose 12 column (Amersham Biosciences). The mobile phase was 20 mM HEPES, 300 mM NaCl, pH 8.0, at a flow rate of 0.5 ml/min. Fractions of 0.5 ml were analyzed by 12% SDS-PAGE followed either by Coomassie Brilliant Blue staining or Western blot analysis using the anti-Myc mAb. The protein concentrations were measured using the Bradford method (Bio-Rad).
Filter Assay for Ligand Binding to ECDs-Various amounts of purified recombinant proteins were incubated for 2 h at 4°C with 25,000 dpm of 125 I-epibatidine (PerkinElmer Life Sciences; specific activity, 2200 Ci/mmol), 20,000 dpm of [ 3 H]epibatidine (PerkinElmer Life Sciences; specific activity, 56.3 Ci/mmol), or 96,150 dpm of [ 3 H]nicotine (PerkinElmer Life Sciences; specific activity, 66.9 Ci/mmol) in a final volume of 50 l of PB-BSA buffer (10 mM phosphate buffer, 0.2% BSA, pH 7.5). The samples were then diluted in 1 ml of 20 mM Tris, 0.05% Triton X-100, pH 7.5, and immediately filtered through an anion exchanger Whatman DE81 filter and presoaked with the same buffer, and the filter was washed four times with 1 ml of the same buffer. The charged DE81 filters bind the free or ligand-bound ECDs, but not free ligands, such as 125 I-epibatidine. In the case of [ 3 H]epibatidine or [ 3 H]nicotine, GF/B glass fiber filters (Whatman) were used and were presoaked for 6 h at 4°C with 1.5% polyethyleneimine, pH 9.0. The cationic polyethyleneimine binds strongly to the glass filters, allowing them to bind acidic proteins and polyanions (31) and consequently free and ligand-bound ECDs, but not the free 3 H-ligands. Bound radioactivity was measured on a ␥or ␤-counter. ⌵onspecific binding was measured in the presence of 1000-fold higher concentration of unlabeled epibatidine or nicotine and the recombinant protein. In all of the experiments, nonspecific binding never exceeded 0.4% of the total used radioactivity.
Saturation Curves Obtained Using Nonlinear Regression-To calculate the dissociation constant (K d ) for 125 I-epibatidine or [ 3 H]nicotine, saturation curve analysis was carried out, according to the model Y ϭ B max *X/(K d ϩ X), where X is the radioligand concentration, and Y is the concentration of specifically bound radioligand. Several concentrations of 125 I-epibatidine or [ 3 H]nicotine were incubated, respectively, for 2 h at 4°C with 0.4 or 40 g, of ECDs in a final volume of 50 l of PB-BSA. Bound radioligands were measured as described above.
Ligand Competition Experiments-Different amounts of unlabeled ligands were added simultaneously with 125 I-epibatidine (25,000 dpm) to 0.4 g of purified recombinant protein in a final volume of 50 l of PB-BSA and the mixture incubated for 2 h at 4°C. The specifically bound radioactivity was then measured by the Whatman DE81 filter assay described above. The inhibition constant (K i ) for each ligand was calculated using the Cheng-Prusoff equation: (32), where IC 50 is the concentration of the ligand that causes 50% inhibition of 125 I-epibatidine binding, determined from the inhibition curve through nonlinear regression according to the model Y ϭ Bottom ϩ (Top-Bottom)/[1 ϩ 10ˆ(X Ϫ LogIC 50 )]. ("Top" is defined as the total binding of labeled ligand in the absence of competitor. "Bottom" is defined as the nonspecific binding of labeled ligand in the presence of a saturating concentration of the competitor.) [ 125 I-epibatidine] is the concentration of 125 I-epibatidine used in the competition experiment, and K d is the dissociation constant for 125 I-epibatidine, estimated as above.
Dynamic Light Scattering Studies-Dynamic light scattering (DLS) analysis of purified recombinant proteins was carried out using a Zetasizer NanoS Instrument (Malvern Instruments) with a helium-neon laser supplying 633-nm light and an output power of 4.0 milliwatt. The samples were placed in a quartz cuvette, and measurements were made at 25°C for an automatically determined time. The results were analyzed using DTS v.4.1 software. The estimated size for each protein molecule is given as the mean hydrodynamic diameter (d⅐nm) of the particles calculated from the intensity of the scattered light, whereas the polydispersity (%) is calculated from the width (nm) of the peak of interest.
In Vitro Deglycosylation of the ␤2-24-␣4-mECD-Approximately 1 mg of purified ␤2-24-␣4-mECD was subjected to deglycosylation under native conditions by the addition of 20,000 units of peptide:N-glycosidase F (PNGase F; New England Biolabs) in 20 mM HEPES, 300 mM NaCl, pH 8 (final volume, 500 l) and incubation at 4°C for 24 h or by the addition of 12,500 units of Endo Hf (New England Biolabs) in 20 mM HEPES, 300 mM NaCl, pH 7, and incubation as above. All of the experiments with deglycosylated ␤2-24-␣4-mECD were performed after removing the endoglycosidases by either gel filtration (PNGase F) or nickel-nitrilotriacetic acid metal affinity chromatography (Endo Hf).

Wild-type and Mutated ECDs-
The singly expressed ␣4and ␤2-ECDs were obtained mainly in the form of microaggregates and multimers with respective apparent molecular masses of 1130 and 480 kDa (data not shown) with a low yield (0.1 and 0.3 mg/liter, respectively). On the basis of our previous studies (28,30,33), we replaced the Cys-loop (Cys-133-Cys-147 in ␣4 and Cys-130 -Cys-144 in ␤2) by the corresponding, more hydrophilic one from the LsAChBP (Cys-123-Cys-136); the expressed mutants were also isolated in the form of microaggregates and multimers, but with a much higher yield (0.3 and 2.5 mg/liter, respectively). More specifically, the ␣4-mECD was mainly eluted in the form of microaggregates and multimers, with respective masses of ϳ1100 and 480 kDa (Fig. 1A), whereas the ␤2-mECD displayed a somewhat better elution profile, because the major fraction eluted as a peak with a mass of about 470 kDa (Fig. 1B).
Coexpression of the Mutated ECDs-We then attempted to coexpress the two mutated ECDs in an effort to obtain ligandbinding pentameric ECD complexes as shown by others using coexpression of intact ␣4 and ␤2 subunits (23,34). The two mECDs were expressed with different tags (␣4-mECD-FLAG and ␤2-mECD-His 6 ) to detect the coexistence of these molecules. On size exclusion chromatography, the elution pattern showed mainly microaggregated proteins and multimers (Fig.  1C). We used two different affinity purification methods for the coexpressed product under native conditions to investigate the formation of complexes between ␣4and ␤2-mECDs. Western blotting of the eluate from an nickel-nitrilotriacetic acid resin using anti-FLAG mAb or of the eluate from an anti-FLAG resin using anti-His 6 mAb (not shown) demonstrated that the coelution of the ␣4and ␤2-mECDs on gel filtration chromatography was probably due to complex formation between these two molecules. However, we did not observe any binding of 125 Iepibatidine to the coexpressed products (Fig. 2, third bar), suggesting that the complexes between the ␣4and ␤2-mECDs were not correctly formed. Because of the insoluble character of the individually expressed wild-type ECDs, probably because of their hydrophobic Cys-loop, we did not attempt to coexpress these forms, considering it unlikely to result in a soluble assembled molecule.

Binding of 125 I-Epibatidine, [ 3 H]Epibatidine, and [ 3 H]Nicotine-
The saturation curves for labeled ligands revealed about an order of magnitude difference between the calculated B max values (in pmol/mg) for the same protein but using the different ligands ( 125 I-or [ 3 H]epibatidine versus [ 3 H]nicotine). This apparent difference may be an artifact caused simply by the fact that the concatamer displays much higher affinity for epibatidine than for nicotine. In fact, the ligand competition experiments (explained below and presented in Fig. 5) show that nicotine can block the binding of practically all 125 I-epibatidine-binding sites.  Competitive Binding of Unlabeled Ligands to the ␤2-24-␣4-mECD Using 125 I-Epibatidine-To further examine the ligand binding properties of the concatamer ␤2-24-␣4-mECD, unlabeled cholinergic ligands were tested in competition experi-ments with 125 I-epibatidine. As shown in Fig. 5A and Table 3, the agonists epibatidine, cytisine, nicotine, acetylcholine, and carbamylcholine exhibited inhibition constants (K i values) of 0.27, 3.24, 20.64, and 242 nM and 2.25 M, respectively, usually approaching the maximum published corresponding values for the intact human ␣4␤2 nAChR. The more selective ␣4␤2 agonists UB-165, RJR-2403, 5-I-A85380, TC-2559, and Ϯ-anatoxin A displayed K i values of 3.18, 33.81, 70.88, and 241 nM and 7.37 M, respectively ( Fig. 5B and Table 3), roughly an order of magnitude lower than the published K i values for the intact ␣4␤2 receptor (38 -40).
Use of Longer ECD Linkers-Assuming that the inability of the ␣4-24-␤2 protein to form the ligand-binding site and confer water solubility was due to too short a linker, we then increased the length of the linker from 24 to 33 residues ((AGS) 11 ), fol-  Table 2.

TABLE 1 DLS analysis of the different mECDs
The measurements were performed using the gel filtration peak fraction of the oligomeric form of the expressed proteins (see legends to Figs. 1 and 2) lowing a construction procedure similar to that for the 24-residue linker constructs for both the ␤2-33-␣4-mECD and ␣4-33-␤2-mECDs. The concatamer ␤2-33-␣4-mECD was shown by size exclusion chromatography to form monomers (data not shown), but its K d for 125 I-epibatidine binding was 1.12 nM, i.e. almost a 3-fold lower affinity than that of the ␤2-24-␣4-mECD dimer. This difference in the dissociation constant is probably due to the linker extension between the mutated ECDs either allowing them to temporarily dissociate or partially preventing ligand access to the binding domain by steric hindrance. Interestingly, the ␣4-33-␤2-mECD construct also eluted mostly as a monomer, in contrast to the microaggregated ␣4-24-␤2-mECD, but did not exhibit any binding of 125 Iepibatidine or [ 3 H]nicotine, similar to the shorter ␣4-24-␤2-mECD (data not shown).
We therefore studied the effect of in vitro deglycosylation of ␤2-24-␣4-mECD on its water solubility and ligand binding properties. We tested two different endoglycosidases: PNGase F and Endo Hf. The ability of these enzymes to deglycosylate the concatamer ␤2-24-␣4-mECDs was analyzed by SDS-PAGE, and the quality of the deglycosylated products was validated by size exclusion chromatography, DLS, and binding affinity for 125 I-epibatidine and [ 3 H]nicotine.
Using PNGase F, the deglycosylated product eluted on size exclusion chromatography as a heterogeneous population of oligomeric molecules (Fig. 3C). SDS-PAGE and Western blot analysis demonstrated the conversion of the heterogeneous smear observed with the glycosylated form into a main sharp band of 55 kDa and a minor band of 57 kDa (Fig. 3C, inset, ϩPNGase), implying that deglycosylation of the ␤2-24-␣4-mECD may be incomplete. DLS analysis illustrated an increase in both size and heterogeneity of the treated protein (Table 1), whereas binding analysis showed a 4-fold decrease in the binding affinity for 125 I-epibatidine and a 2-fold decrease in the binding affinity for [ 3 H]nicotine ( Fig. 4B and Table 2).
In contrast, when Endo Hf was used, the deglycosylated concatamer eluted mostly as a monomer (Fig. 3C). Moreover, Western blot analysis showed that deglycosylation converted the heterogeneous smear seen with the glycosylated form (Fig. 3B) into a single major band, which migrated with a molecular mass of ϳ55 kDa (Fig. 3C, inset, ϩEndoH). DLS analysis confirmed the smaller size of the deglycosylated concatamer, the hydrodynamic diameter being 9.4 nm compared with 12.5 nm for the glycosylated molecule (Table  1). Finally, the deglycosylated molecule fully retained its 125 Iepibatidine and [ 3 H]nicotine binding abilities (Fig. 4C), FIGURE 5. Competitive inhibition of 125 I-epibatidine binding to ␤2-24-␣4-mECD by unlabeled ligands. Inhibition of 125 I-epibatidine binding to ␤2-24-␣4-mECD was assessed by coincubation competition experiments with unlabeled ligands, as described under "Experimental Procedures." The ability of these ligands to inhibit 125 I-epibatidine-binding to ␤2-24-␣4-mECD was determined by the IC 50 value obtained from the curves and expressed as the K i (values given in Table 3).

DISCUSSION
Human ␣4␤2 nAChR is of great importance because of its implication in various physiological functions and pathological conditions. We therefore studied the expression of the ␣4and ␤2-ECDs using the eukaryotic expression system P. pastoris to achieve a high yield and good solubility of proteins with intact ligand-binding sites.
Several approaches were used before we finally obtained a recombinant molecule with the required characteristics. The individual wild-type ECDs were expressed mostly as microaggregates with a poor expression yield (data not shown). Next, we expressed mutated ␣4and ␤2-ECDs, in which the hydrophobic region of the Cys-loop domain was replaced by the corresponding, more hydrophilic region of the LsAChBP. We have used this replacement previously in the expression of the human ␣7-ECD, and this led to a significant improvement in the solubility and ligand binding properties of the recombinant molecule (28,30,33). This mutation again led to a significant increase in the expression yield of the singly expressed ECDs but not to a large improvement in their solubility (Fig. 1, A and B). No 125 I-epibatidine binding to the singly expressed ␣4or ␤2-ECD was observed, as expected because of the fact that the ligand-binding site is formed at the interphase of the two subunits.
We then tried to coexpress the ␣4and ␤2-mECD to obtain functional (i.e. ligand binding) complexes, as shown previously by coexpression of intact ␣4 and ␤2 subunits (41) or ECDs carrying the transmembrane M1 domain (23). Gel filtration chromatography demonstrated that the coexpressed polypeptides existed mostly as microaggregated complexes (Fig. 1C). However, when these recombinant proteins were subjected to binding experiments with 125 I-epibatidine, no specific binding was detected, implying the lack of a properly assembled binding site (Fig. 2).
These results led us to try a different strategy of expressing the ␣4-mECD and ␤2-mECD linked via a synthetic neutral flexible peptide (AGS) 8 in an effort to facilitate the proper assembly of these two mutant ECDs. For the first time, we observed that the glycosylated concatamer ␤2-24-␣4-mECD was expressed with a molecular mass corresponding to a dimeric, rather than an oligomeric, form and in the absence of microaggregates (Fig.  3B). More importantly, this protein exhibited ligand binding properties approaching those of the native ␣4␤2 receptor for both 125 I-epibatidine and [ 3 H]nicotine (Table 2). Additionally, competition experiments between 125 I-epibatidine and several unlabeled ligands showed that the K i values for these ligands approach those reported for the intact ␣4␤2 receptor (Fig. 5 and Table 3). In particular, the K i values of the less specific ligands (epibatidine, cytisine, nicotine, acetylcholine, and carbamylcholine) are only about two times higher than the highest published K i values for the intact AChR, and those of the more specific ␣4␤2 ligands (UB-165, RJR-2403 and Ϯ-anatoxin A), with the exception of 5-I-A85380, are about an order of magnitude higher than the larger published K i values of the intact AChR. In contrast, the ␣4-24-␤2-mECD concatamer was obtained in the form of microaggregates and in small amounts, as shown in Fig. 3A, and displayed no binding activity for either 125 I-epibatidine (Fig. 2)

or [ 3 H]nicotine (data not shown).
We also tested the use of a longer linker for both ECD combinations, but no improvement in their pharmacological characteristics was observed. In fact, the binding affinity of the ␤2-33-␣4 construct was lower than that observed for the ␤2-24-␣4 construct, whereas the ␣4-33-␤2 construct, although it appeared much more soluble than the ␣4-24-␤2 construct, still did not bind ligands.
In preliminary experiments, we also tried to generate pentameric ECDs. We first coexpressed each of the two concatamers (␣4-24-␤2 and ␤2-24-␣4) bearing the His 6 tag with either the ␣4or ␤2-mECD bearing the FLAG tag. No 125 I-epibatidine binding was observed with the products of coexpressions involving ␣4-24-␤2 (with either the ␣4or ␤2-mECD), and we did not characterize them further. Only the coexpression of ␤2-24-␣4 mECD with ␤2-mECD-FLAG resulted, as expected, in 125 I-epibatidine binding complexes, but DLS studies revealed extensive size heterogeneity of the isolated complexes, precluding any significant structural studies on these molecules, at least at this stage.
Finally, we also constructed two concatenated mECD trimers (␤2-24-␤2-24-␣4 and ␤2-24-␣4-24-␤2). Unfortunately, SDS-PAGE and Western blot analysis revealed the presence of a mixture of truncated proteins corresponding to trimeric, dimeric, and monomeric ECDs, suggesting that the P. pastoris expression system that we use, although appropriate for the expression of functional dimeric ECDs, is not suitable for the expression of more complicated constructs, probably because of intrinsic recombination at the plasmid level prior to integration in the yeast genome. Similarly, the coexpression of these trimers with ␣4and ␤2-mECDs (all four combinations) did not improve the characteristics of the expressed proteins.
It is well established that glycosylated proteins usually withstand the production of diffraction quality crystals, probably because of the high flexibility and heterogeneity of the polysaccharide chains. Therefore, it was important to know whether deglycosylation of ␤2-24-␣4-mECD would retain its functional and water solubility properties or not. Although glycosylation seems necessary for the binding of ␣-bungarotoxin to muscle type nAChR (29,42), it does not have any effect on ␣-bungarotoxin binding to recombinant human ␣7 ECDs (28,30), reflecting a differential role of glycosylation in protein folding between the different nAChR subtypes. Thus, it was unclear whether glycosylation of the ␣4 and ␤2 subunits had a significant effect on the pharmacological properties of this receptor subtype. We therefore attempted to deglycosylate in vitro the concatamer ␤2-24-␣4-mECD, under native conditions to determine (a) the role of glycosylation in the ligand-binding site and (b) whether the deglycosylated protein has the characteristics needed for crystallization efforts. We tested two different endoglycosidases: PNGase F and Endo Hf. Endo Hf cleaves within the chitobiose core of high mannose and some hybrid oligosaccharides on N-linked glycoproteins, leaving a residue of GlcNAc linked to asparagine, whereas PNGase F cleaves between the innermost GlcNAc and asparagine residues of high mannose, hybrid, and complex oligosaccharides on N-linked glycoproteins, leaving no saccharide residues but converting the asparagine to aspartate (43).
In the case of PNGase F, the treated concatamer was not isolated as a completely deglycosylated protein. Moreover, the final product was a heterogeneous mix of proteins in different glycosylated states and with reduced binding affinity for 125 Iepibatidine, forming higher order oligomers (Figs. 3C and 4B and Table 2). In contrast, the results were encouraging when Endo Hf was used for in vitro deglycosylation of the ␤2-24-␣4-mECD concatamer. Size exclusion chromatography revealed the formation of a consistently deglycosylated protein, probably in the monomeric form (Fig. 3C), giving by far the best DLS results in terms of polydispersity and molecular size (Table 1). Interestingly, it was observed that the binding site was intact, because the deglycosylated protein retained its binding affinity for 125 I-epibatidine ( Fig. 4C and Table 2). We assume that this difference between the two endoglycosidases is due to the remaining residue of GlcNAc, which stabilizes ␤2-24-␣4-mECD. Moreover, the observation that the deglycosylated form binds 125 I-epibatidine to the same extent as the glycosylated form may be very useful in crystallization studies, because it means that the deglycosylated form represents a good candidate for use in structural studies. In addition, cocrystallization of the deglycosylated molecule with the ligands may increase the chance of achieving diffraction quality crystals and will allow the detailed study of ligand-receptor interactions.
In conclusion, although coexpression of the ␣4and ␤2-mECDs of the human ␣4␤2 nAChR did not result in the assembly of the ligand-binding site, linking of the mECDs with a flexible linker in a specific sequence (␤2-24/33-␣4) resulted in a water-soluble molecule with very good ligand binding properties. Furthermore, its deglycosylation showed that the carbohydrate arms do not contribute to the ligand-binding site of the ␣4␤2 receptor. Thus, the ␤2-24-␣4-mECD concatamer, with its very satisfactory yield and ligand binding and biochemical properties, seems a promising molecule (in both the glycosylated and deglycosylated forms) for structural studies, essential for rational drug design to treat relevant diseases.